Create README.md

README.md

@@ -0,0 +1,39 @@
+## Task Description
+The task involves training a model to evaluate two pieces of text. One of the texts has been subtly augmented by a LLM (specifically, the larger 14B variant of my [corruption models](https://huggingface.co/Quest-AI/quest-corruption-7b-s375-v3-GRPO)). The model must provide notes and a subsequent judgment afterwards in consecutive XML tags.
+
+![GRPO Task Overview](https://files.catbox.moe/r5dm86.png)
+
+## Example Format
+
+The base model is provided a system prompt that establishes the expected template, as well as two randomly ordered A/B samples containing "real" vs "synthetic" samples for the input:
+
+```
+REQUEST: You are to judge the better of the two samples and determine which of the following samples is better using a short judgement that is no longer than (and no shorter than) exactly 128 tokens.
+
+Respond with an exactly 128 tokens tag labeled <notes> that contains your notes, and then <judgement> which is just the letter that you are picking.
+
+For example:
+
+JUDGE: <notes>
+Sample A is superior to Sample B... (example notes)
+</notes>
+<judgement>A</judgement>
+
+Now, it is your turn.
+
+[Sample A]:
+Included is a pre-test, post-test, and vocabulary quiz on the 8th grade math standard functions (8.F). 1.) Determine if a graph represents a function 2.) State the domain and range of a relation 3.) Plot points on a graph to determine if the table represents a function 4.) State if a function is decreasing, increasing, or constant 5.) Determine the output of a function machine 6.) Determine the recursive and explicit equation 7.) Determine the minimum, maximum, increasing interval, and decreasing interval of a graph 8.) Determine the rate of change, initial value, independent value, and dependent variable given a graph 9.) Sketch a graph given a situation The vocabulary included is dependent, output, function, domain, range, decreasing function, input, range, non-linear function, relation, increasing function, and function notation. Total Pages: 9 (18 including answer key) Answer Key: Included Document File: PDF
+
+[Sample B]:
+Included is a pre-test, post-test, and vocabulary quiz on the 8th grade math standard functions (8.F). 1.) Determine if a graph represents a function 2.) State the domain and range of a relation 3.) Plot points on a graph to determine if the table represents a function 4.) State if a function is increasing, decreasing, or constant 5.) Determine the output of a function given 6.) Determine the input of a function given 7.) Determine a function rule given ordered pairs or a table of values. 8.) Graph functions using a table of values and determine a trend line in a graph 9.) Write a data table situation The vocabulary included is dependent, output, function, domain, range, decreasing function, input, range, non-linear function, relation, increasing function, and function notation. Total Pages: 9 (18 including answer key) Answer Key: Included Document File: PDF
+
+JUDGE:
+```
+
+A correct output follows this structure:
+```
+<notes>
+Sample A provides more specific and thoroughly defined tasks. It mentions "function machine," "recursive and explicit equation," and detailed graph analysis with "minimum, maximum" and intervals. Sample B contains incomplete phrases like "output of a function given" without completing the thought, making it less coherent and precise than Sample A.
+</notes>
+<judgement>A</judgement>
+```